Developing a prediction model for the risk of dissociative psychopathology from trauma and trait responsiveness to verbal suggestion

This study developed and validated a prediction model using elastic net logistic regression that integrates developmental trauma and trait responsiveness to verbal suggestion to effectively identify individuals at risk for dissociative psychopathology, achieving an AUROC of .77.

The Gist

Imagine your mind as a complex house with many rooms. Sometimes, when things get too overwhelming, the house has a special safety mechanism: it locks certain doors or dims the lights in specific rooms to protect the people inside. In psychology, this is called dissociation. It's like the mind hitting a "pause" button or creating a foggy barrier between a person and their thoughts, memories, or feelings. While this can happen to anyone, for some, it becomes a persistent problem that makes daily life very difficult.

This paper is like a team of detectives trying to build a better early-warning system to spot who might be at risk of living in this "foggy house" before the problem gets too severe.

The Two Main Clues

The researchers focused on two main clues that might predict who is at risk:

1. The "Storm" (Trauma): We already know that experiencing bad things in life, especially when growing up (like abuse or neglect), is a major factor. Think of this as a severe storm battering the house. The more intense the storm, the more likely the house is to need those safety locks.
2. The "Sponge" (Suggestibility): This is the new, interesting clue. Some people are naturally more like sponges when it comes to verbal suggestions. If someone says, "Your arm feels heavy like a lead weight," a "sponge" person might actually feel their arm get heavy, even if they know it's just words. The researchers call this Responsiveness to Verbal Suggestion (REVS). It's not about being easily tricked; it's about how easily your mind can be guided to change its perception.

The big question the team asked was: Do these two clues work together? Does having a "sponge-like" mind make the "storm" of trauma hit harder, leading to more dissociation?

The Experiment: A Digital Simulation

To find out, the researchers gathered data from over 1,100 people who were not currently in therapy (a "non-clinical" group). They asked these people three things:

• How much "storm" they faced: They filled out a checklist of traumatic events.
• How "sponge-like" they are: They played a computer game where they had to follow verbal instructions (like "your hand is stuck to the table") and rated how involuntary or automatic those feelings felt.
• How much "fog" they experience: They answered questions about how often they felt disconnected from themselves or their memories.

Using a sophisticated computer algorithm (think of it as a very smart sieve that filters out noise to find the true signal), they tried to build a model that could predict who was in the "at-risk" group.

What They Found

The results were like finding a specific pattern in a puzzle:

• The "At-Risk" Group: About 7% of the people were flagged as high-risk. These individuals tended to be younger, had experienced more trauma, and were much more "sponge-like" (highly responsive to suggestions).
• The Power of the "Sponge": Surprisingly, the "sponge" factor (suggestibility) was actually a stronger clue than the trauma itself. Specifically, the feeling that the suggestions were involuntary (like "I couldn't stop my arm from moving even if I wanted to") was a huge red flag.
• The Combination: The best prediction came from looking at both the storm and the sponge together. The model suggested that when a person who is naturally a "sponge" faces a "storm," the risk of developing a foggy, dissociative mind goes up.
• Age Matters: The model also noticed that younger people were more likely to be in the at-risk group. As people get older, the risk seems to drop, perhaps because they develop better coping mechanisms or the "storm" effects fade over time.

How Good Was the Prediction?

The researchers tested their model to see how accurate it was. Imagine a security guard trying to spot intruders:

• The model was 77% accurate overall.
• It was very good at catching the people who were at risk (it didn't miss many of them).
• However, it wasn't perfect at ruling out people who weren't at risk (it sometimes sounded the alarm for people who were actually fine).

Think of it like a smoke detector that is very sensitive: it might beep a few times when you just burnt toast (false alarm), but it's excellent at making sure you don't miss a real fire.

What This Means (and What It Doesn't)

The paper concludes that we shouldn't just look at trauma to understand dissociation. We also need to look at how a person's mind naturally responds to suggestions. It's like saying that to understand why a house burns down, you need to know both about the fire (trauma) and how flammable the building materials are (suggestibility).

Important Note: The authors are very careful to say this is a research tool, not a medical diagnosis kit. They built this model using a computer program on a group of regular people, not patients in a hospital. They are not saying doctors should use this right now to diagnose patients. Instead, they are saying, "Hey, we found a pattern that works pretty well in a simulation. If we test this more with real patients and refine it, it could one day help doctors spot people who need help earlier."

In short, the paper suggests that the key to understanding who might struggle with dissociation isn't just the bad things that happened to them, but also the unique way their mind is wired to process the world around them.

Technical Summary

Technical Summary: Predictive Modelling of Dissociative Psychopathology from Trauma and Trait Responsiveness to Verbal Suggestion

Problem Statement
Dissociative disorders (DDs) and associated symptoms are frequently under- and misdiagnosed, leading to delayed treatment, excessive patient suffering, and increased risks of comorbidity and suicidality. While developmental trauma is widely recognized as a primary antecedent of dissociative psychopathology, the "trauma model" often neglects the role of trait responsiveness to verbal suggestion (REVS). Conversely, REVS has been linked to dissociation but is rarely studied in conjunction with trauma in a predictive framework. Furthermore, existing prediction models often focus on trauma alone or fail to integrate the complex factor structure of suggestibility (e.g., behavioural responses vs. the subjective sense of involuntariness). This study addresses the gap in developing a multivariate prediction model that integrates developmental trauma, trait REVS, and demographic variables to identify individuals at risk for dissociative psychopathology within a non-clinical population.

Methodology
The study utilized a pooled dataset of 1,104 non-clinical participants drawn from four previous cross-sectional studies. The methodology involved three primary phases:

1. Measurement and Factor Analysis:

   • Trauma: Assessed via the Trauma Experiences Checklist (TEC). Scores were transformed into percentages to handle missing data in one subset.
   • Dissociation (Outcome): Measured using the Dissociative Experiences Scale (DES). The primary outcome was a binary classification of "at-risk" status based on the probability of membership in the DES taxon (DES-T), using a conservative threshold of $p \geq 0.95$. A secondary analysis used a total DES score threshold of >30.
   • REVS: Assessed using the Brief Suggestibility Scale (BSS), a computerized measure of non-hypnotic direct verbal suggestibility. The BSS captures both behavioural responses and perceived involuntariness. To address the complex factor structure of REVS, the authors fitted bifactor models to the BSS data. This yielded a general behavioural factor (BSS:B-G), a general involuntariness factor (BSS:I-G), and a specific perceptual involuntariness factor (BSS:I-PF).

2. Prediction Modelling:

   • An elastic net logistic regression model was employed to predict the binary risk status. This method combines ridge and LASSO penalties to handle multicollinearity and select the most predictive variables.
   • Predictors: Included the BSS composite (BSS-C), the derived bifactor scores (BSS:B-G, BSS:I-G, BSS:I-PF), TEC scores, age, and interaction terms (e.g., Age × TEC, BSS-C × TEC).
   • Validation: The model was internally validated using 10-fold cross-validation.

3. Evaluation Metrics:

   • Model performance was assessed using the Area Under the Receiver Operating Characteristic Curve (AUROC), sensitivity, specificity, Youden's Index, and Brier scores.

Key Results

• Sample Characteristics: 7% ($n=77$) of the sample was classified as at-risk for dissociative psychopathology. This group was significantly younger, reported higher levels of trauma, and exhibited elevated scores across all REVS and dissociation measures compared to the non-risk group.
• Factor Structure: The bifactor analysis confirmed that a single general factor best explained the behavioural subscale, while the involuntariness subscale required a two-factor solution (general and perceptual).
• Model Performance:
  • The elastic net model retained 9 of the 10 predictors.
  • The model achieved an AUROC of 0.77 (95% CI: 0.73–0.82), indicating good discrimination.
  • Sensitivity was 78%, while specificity was 67%.
  • The Brier score was 0.06, reflecting excellent model fit, with the predictor set explaining 75% of the mean squared error.
• Predictor Significance:
  • Age was the only variable reaching statistical significance in the final model (negative coefficient, indicating risk decreases with age).
  • Trauma (TEC) and REVS measures (BSS-C, BSS:I-PF) were retained in the model but were not individually significant, likely due to high multicollinearity and the shrinkage properties of the elastic net.
  • Interaction terms (e.g., Age × TEC, BSS-C × TEC) were retained but showed weak effects.
  • Notably, the effect size for the BSS-C (incorporating involuntariness) was larger than that for trauma, suggesting the added value of assessing the sense of agency.
• Robustness: A secondary analysis using a total DES score threshold (>30) yielded nearly identical results (AUROC = 0.762), confirming the model's stability across different definitions of at-risk status.

Key Contributions

1. Integration of Trauma and REVS: The study demonstrates that developmental trauma and trait REVS can be successfully integrated into a single predictive framework, supporting the Diathesis-Stress model where suggestibility may moderate the impact of trauma.
2. Inclusion of Involuntariness: By utilizing bifactor modeling, the study highlights that the perceived involuntariness of responses to suggestions is a critical component of REVS in the context of dissociation, offering a more nuanced predictor than behavioural response alone.
3. Multivariate Prediction: The paper moves beyond univariate associations to provide a multivariate model that identifies at-risk individuals with moderate accuracy (77% classification probability), addressing the need for better early identification tools.
4. Methodological Rigor: The use of an automated, standardized measure of suggestibility (BSS) circumvents experimenter effects, and the use of elastic net regression allows for variable selection in the presence of correlated predictors.

Significance and Claims
The authors claim that these findings reinforce trauma and trait REVS as risk factors for dissociative psychopathology and demonstrate that they can be combined to identify at-risk individuals. The study suggests that risk for dissociative psychopathology is not solely driven by trauma but reflects a constellation of cognitive characteristics (specifically altered sense of agency/involuntariness) and demographic factors.

The paper maintains a modest tone regarding clinical application. It explicitly states that the model is currently based on a non-clinical sample and lacks external validation; therefore, it should not be used for clinical diagnosis or to guide practice at this stage. The authors posit that the model lays a foundation for future research and potential early identification strategies, provided that independent replication, the inclusion of refined trauma measures, and validation against clinical interviews (e.g., SCID-D) are conducted. The work challenges the view that dissociation is unrelated to suggestibility and argues against reducing the link between REVS and dissociation solely to issues of iatrogenesis or false memory.